CORE Metadata, citation and similar papers at core.ac.uk Provided by Elsevier - Publisher Connector NeuroImage 55 (2011) 732–741 Contents lists available at ScienceDirect NeuroImage journal homepage: www.elsevier.com/locate/ynimg The functional neuroanatomy of morphology in language production Dirk Koester a,b,⁎, Niels O. Schiller c,d,e a Center of Excellence – Cognitive Interaction Technology (CITEC), Bielefeld, Germany b Bielefeld University, Faculty of Psychology and Sport Science, Bielefeld, Germany c Leiden Institute for Brain and Cognition, Leiden, The Netherlands d Leiden University Centre for Linguistics, Leiden, The Netherlands e Leiden University, Faculty of Social and Behavioural Sciences, Leiden, The Netherlands article info abstract Article history: The present study investigated the neural correlates of morphological priming in overt Dutch language Received 16 July 2010 production using a long-lag priming paradigm. Compound words were read out loud as primes that were Revised 4 November 2010 morphologically related to picture names (e.g. the word jaszak, ‘coat pocket’ was used for a picture of a coat; Accepted 12 November 2010 Dutch jas), or primes were form-related, but not morphologically related monomorphemic words (e.g. Available online 23 November 2010 jasmijn, ‘jasmine’). The morphologically related compounds could be semantically transparent (e.g. eksternest, ‘magpie nest’) or opaque (e.g. eksteroog, lit. ‘magpie eye,’‘corn,’ for a picture of a magpie, Dutch ekster). These Keywords: Language production four priming conditions were complemented by two matched, unrelated conditions. The production of Overt speech morphologically related, complex words but not the production of form-related words facilitated subsequent Priming picture naming. Also, morphologically related but not form-related words led to a neural priming effect in the Morphology left inferior frontal gyrus (LIFG). The effects did not differ for transparent and opaque relations. The results Mental lexicon point to a functional role of LIFG in morphological information processing during language production Semantics contrary to previous meta-analytic findings. Specifically, morphological priming effects in language BA 47 production seem to be independent from semantic overlap. However, further research should confirm the LIFG independence of morphological and phonological factors. It is suggested that LIFG subserves word form fMRI encoding in language production. © 2010 Elsevier Inc. Open access under the Elsevier OA license. Introduction Compound words (e.g. coat+pocket) are concatenations of free morphemes and have an internal structure. In the present study, such Language is a characteristic feature of all humans (Hockett, 1960) compounds were used to prime the process of naming pictures in and comprises largely of structured sequences of words, the building Dutch, an Indo-European language. The internal structure is hierar- blocks of sentences. Words, in turn, comprise often times of chical in that one constituent morpheme, the last in Dutch, structured sequences of morphemes. The former regularities are determines the morphosyntactic features and the semantic category part of syntax; the latter are described in morphology, which deals of the whole compound (Booij, 2002a; Downing, 1977). The crucial with the internal structure of words. question was whether and where in the human brain morphological Recognizing the structure of morphologically complex words is overlap between the compound words and the picture names would essential for meaning construction. For example, when hearing the affect picture naming. To our knowledge, there is no study word “worthless,” the listener can apply the morphological structure investigating morphological compound processing in speaking by adjective+suffix to understand “worthless” as indicating an entity of means of functional magnetic resonance imaging (fMRI). no significant value. In contrast, when hearing “worth less” (e.g. in a Language production is thought to involve a sequence of four comparative sentence; “X is worth less than Y”), the listener should major cognitive processes: conceptual preparation, lexical access, not apply the same morphological structure but interpret the phonological processing, and articulation (Caramazza, 1997; Dell, utterance as an expression of relative value referring to two entities 1986; Garrett, 1988; Levelt, 1989; Levelt et al., 1999). Conceptual denoted by X and Y. preparation refers to the activation of the concepts that relate to the Here, we are interested in the functional neuroanatomical intended utterance—when naming a picture, the concept of the correlates of morphological priming in overt language production. depicted object. The conceptual activation spreads to associated lexical representations. Next, word form encoding follows, i.e. phonological information needed for pronunciation is processed. ⁎ Corresponding author. Center of Excellence – Cognitive Interaction Technology, fi Bielefeld University, 33615 Bielefeld, Germany. Fax: +49 521 106 24 20. The phonological word form is nally used for articulation by calling E-mail address: [email protected] (D. Koester). up the corresponding gestural scores. For the production of 1053-8119 © 2010 Elsevier Inc. Open access under the Elsevier OA license. doi:10.1016/j.neuroimage.2010.11.044 D. Koester, N.O. Schiller / NeuroImage 55 (2011) 732–741 733 morphologically complex words, it has been suggested that these results, it has been suggested that morphology emerges from the words are not planned as one chunk. Rather, they are planned convergence of form and meaning processing (Devlin et al., 2004; see serially, one morpheme after the other (Roelofs, 1996; Levelt et al., also Joanisse and Seidenberg, 2005 for inflectional morphology). 1999). This construction of complex words from separately stored Others found no effect when comparing the reading of morpholog- morphemes (or word stems in the case of compounds) is called the ically simple and complex words, besides a word class effect (verbs vs. decompositional view of word production, and it has been proposed nouns/adjectives; Davis et al., 2004). Such results argue against an to be purely morphological, that is, without a contribution of independent representation of morphological information in reading semantic information (Roelofs and Baayen, 2002). In contrast to (but see Gold and Rastle, 2007). the decompositional view, other researchers proposed that morpho- Others investigated neural correlates of morphological priming in logically complex words are not composed from their constituent unmasked single word reading in English (Bozic et al., 2007). These morphemes during speaking, but rather stored and retrieved as authors used a long-lag priming paradigm in which the prime word whole units (so-called full-listing hypothesis; Butterworth, 1983; preceded the target word by a number of intervening trials (see Bybee, 1995; Janssen et al., 2008). below). An enhanced neural response was found in the left inferior Even though details of the architecture of the language production frontal gyrus (LIFG; BA 47) but no effect in the left occipito-temporal system are still debated (e.g. to what extent activation flow is cortex when comparing the hemodynamic response for morpholog- cascading rather than discrete; Damian and Bowers, 2003; Jescheniak ically complex with simple words. Here, the neural activation for the et al., 2002, 2009; Levelt, 2001; Morsella and Miozzo, 2002; Navarrete morphologically complex words did not differ between semantically and Costa, 2005; Roelofs, 2003), it should be noted that morphological transparent (e.g. hunter) and semantically opaque words (e.g. encoding follows lexical selection (Levelt et al., 1999). Accordingly, corner).1 However, when comparing the hemodynamic response for morphological structure has been allocated at the word form level the prime and the target words, Bozic et al. (2007) found reduced (Zwitserlood et al., 2000). In terms of functional neuroanatomical activation in LIFG only for semantically transparent (e.g. hunter– localization, morphological information has been suggested to play a hunt) and opaque (e.g. corn–corner) pairs which were proposed to be role in the first sub-stage of word form encoding (Indefrey and Levelt, morphologically complex but no difference was found for form- 2004). related (e.g. scandal–scan) or purely semantically related pairs (e.g. In a recent meta-analysis of behavioral and fMRI studies of accuse–blame). Consistent with these results, Meinzer et al. (2009) language production, Indefrey and Levelt (2004) assessed the brain reported an involvement of the left BA 47 for the reading of German areas functionally associated with different cognitive processing derivations that are morphologically complex. Interestingly, these stages in language production. However, the neuroanatomical authors reported an increased activity in the LIFG and left posterior correlates of morphological priming remain controversial. Based on middle temporal gyrus (MTG) for more complex vs. less complex this meta-analysis, phonological code retrieval has been localized in derivations. Generally, these data are consistent with a role of the LIFG the left posterior superior and middle temporal gyri (so-called in morphology and morphosyntax (Marangolo et al., 2006; Marslen- Wernicke's area together with the right supplementary motor area Wilson and Tyler, 1998; Tyler et al., 2002, 2004; Ullman, 2001; Laine